Image Texture Feature Extraction Based on Gabor Transform

As a regional feature, texture is a description of the spatial distribution of the image pixels. As texture can fully utilize the image information, it can become an important basis to describe and recognize the image. Compared with other image features, texture can take both the macro image properties and micro structure into consideration; therefore, feature has become a significant feature to be extracted in the target recognition. This paper analyzes the performance of Gabor filter, designs a multi-channel Gabor filter and selects its parameters. By extracting various local feature information (direction, phase, energy and so on) of the image with multichannel filter, this method can decompose an image into the sub-images based on different frequency directions and channels and it classifies the sub-images with the statistical feature extracted from the sub-image as the texture feature to represent different texture region features of the image. It is quite convenient to realize and it can obtain the optimal comparison effect in the spatial and frequency domains. The experiment result shows that the Gabor filter of this paper have excellent performance in analyzing the frequency and the direction information of the local regions of digital image.